The Hybrid Feature Selection Algorithm Based on Maximum Minimum Backward Selection Search Strategy for Liver Tissue...

Area of Science:

• Medical Imaging
• Computer Science
• Bioinformatics

Background:

• Accurate classification of liver tissue pathological images is crucial for diagnosis.
• Existing feature selection methods may struggle with unbalanced datasets and limited feature evaluation.

Purpose of the Study:

• To develop a novel and efficient feature selection algorithm for liver tissue pathological image classification.
• To address the limitations of existing evaluation criteria and search strategies in handling unbalanced samples.

Main Methods:

• A hybrid feature selection approach combining rough and precise selection stages.
• Introduction of Maximum Minimum Backward Selection (MMBS), a heuristic search algorithm.
• Development of Weighted Discernibility of Feature Subsets (WDFS) as an evaluation strategy for MMBS, suitable for unbalanced samples.

Main Results:

• The proposed MMBS algorithm effectively identifies optimal feature subsets.
• The WDFS evaluation criteria demonstrate robustness with unbalanced liver tissue pathological image datasets.
• The hybrid feature selection algorithm achieved good classification performance.

Conclusions:

• The novel feature selection algorithm, incorporating MMBS and WDFS, significantly improves liver tissue pathological image classification.
• This approach offers a robust solution for handling unbalanced data in medical image analysis.
• The method shows promise for enhancing diagnostic accuracy in liver pathology.